Introductions of My Jazz Habits: Different playlist for different moods

When I am studying or just relaxing I often listen to Jazz. It can set a mood and is often good background music whilst focusing on other things. I have got two main spotify jazz lists I listen to.

One of the lists is an arbitrary mix of different Jazz songs I enjoy listening to. The songs are from different Jazz sub-genres and were composed at different times. To me, this playlist would describe a good general overview of my jazz listening habits. Therefore, I call this my General Jazz list.

The other jazz list consists of more “toned-down” versions of popular jazz songs. These songs are made by a select set of artist to be used as smooth background music and to set a relaxed atmosphere. There are no vocals and abrupt changes in the song “feel” less intense. To me, this playlist always gives feels like a more jazzy version of elevator music. Thus, I call this my Elevator Jazz list. I often use this list as background music in social settings, but I mostly listen to this playlist when I am studying. For example, during the creation of this portfolio :)

There is a slight overlap between the two lists. For example, the classic jazz song “Take five - The Dave Bruback Quartet” is part of general jazz list, whilst a different version “Take 5 - New York Jazz Lounge” is part of the elevator jazz list. These songs are extremely similar, yet sometimes I want to listen to the normal jazz list and sometimes to the elevator jazz list.

Given that the elevator jazz list consists of simplified versions of popular jazz songs, it would be interesting to find out how they differ. Thus I have created a version of the elevator jazz list with the original tracks. I call this playlist the Elevator Jazz (Originals). These list will also be compared.


All this leads into multiple interesting things that can be researched:


To answer these questions, first some exploratory visualizations will be done. Then a combination of different feature analysis and clustering will be performed to draw conclusions.

Exploring the Corpus: The biggest differences between Elevator jazz and General jazz

After exploring the two lists, some notable differences and similarities were found. First of all, the elevator jazz list has lower energy and a higher acousticness. This seems to be in line with what you could expect from a list that can be used as smooth background music.

Secondly, it was found that the elevator jazz list has a lower speechiness. This is also expected, because the list was designed to not contain any vocals.

To the contrary, the elevator jazz list has a higher danceability, this is unexpected because the list is mainly listened to when I am studying. This is also interesting because this list has a lower overal energy, yet a higher danceability.

Exploring the Corpus: A track-level feature overview of the means of Elevator jazz and General jazz

Here, we see how the mean of the track-level features are different between the elevator jazz playlist and the original jazz playlist. It is notable that the standard deviations of the different features (as indicated by the size of the dots) of the elevator list are on average lower than the standard deviations of the general list. This tells us that the songs have more similar features in a list that consists of simple jazz songs than in a jazz list that consist of random jazz songs from different subgenres.

The original list has a slightly higher speechiness and liveness, but really averages out on the other features. This might implicate that a lot jazz (at least the jazz that I listen to) have a low speechiness and liveness in general. And that the other features might be somewhat fixed within a subgenre, but average out over my complete jazz listening preferences.

Exploring the Corpus: A track-level feature overview of the means of Elevator jazz and Original jazz

Now lets dive into the elevator jazz list, and explore how these remade songs differ from their originals. Most of the features are similar in value and standard deviation. Therefore it is hard to draw any conclusions from this, other than “two only slighlty different versions of the same song have similar features.”

This is not very helpful, but there is one clear outlier. The instrumentalness is notably higher in the elevator jazz playlist. It might be usefull to dive into spotifys definition of instrumentalness to explain this:

“Instrumentalness predicts whether a track contains no vocals. “Ooh” and “aah” sounds are treated as instrumental in this context. Rap or spoken word tracks are clearly “vocal”. The closer the instrumentalness value is to 1.0, the greater likelihood the track contains no vocal content. Values above 0.5 are intended to represent instrumental tracks, but confidence is higher as the value approaches 1.0."


Thus, instrumentalness is a predictor for whether or not a track contains vocals. From my own experience (which is many many hours of listening to this playlist on repeat) the whole elevator jazz playlist contains absolutely zero vocals. To the contrary, many of the original songs do. This explains the difference in instrumentalness.

Further Exploration of the three playlists: instrumentalness vs. energy

Now that we know that the instrumentalness can be used to distinguish between the lists, we can use this to make a bigger plot with all three the playlist. Differences are most notable when instrumentalness is being displayed against energy, a slight jitter is added to deal with overlapping points. This plot might show some indication of some slight clustering:


Elevator Jazz: Most of the songs in this playlist are found in the top-left corner. As established before, these songs hava a high instrumentalness and a mostly low energy.

Elevator Jazz (originals): The bottom-left corner contains the majority of the songs. Apparently most of these songs have vocals, and thus a very low or no instrumentalness. Furthermore, most of these songs have a energy of less than 0.5.

General Jazz: This list seems to be the hardest to characterize, which makes sense given that the list is more random. But a decent majority of the songs seem to have an energy of more than 0.5.

Different Timbre Coefficients between the Elevator Jazz and its Original Jazz List

Here you can see a summary of the different timbre coefficients between “Elevator Jazz” and "Elevator Jazz (Originals).

Outliers within the corpus: confusing relations between the Tempo of different Songs

Lets have a look at the outlier present in the “Elevator Jazz” list. We can see a boxplot of the differences in tempo for the “elevator jazz” list and the list of the original songs of that list (upper left). As you can see they have a mostly similar distribution. But the “Elevator Jazz” list has one outlier: “I wanna Be Around - Beegie Adair” at an average tempo of 215.908 BPM. This seems very interesting. Thus it is time to compare this song with its original “I wanna Be Around - Tony Bennet”, and a song in the playlist with an average tempo: “You Are the Sunshine of My Life - New york Bar Quartett”.

After listening to these three songs it is apparent that both song [B] and [C] actually have a higher tempo than song [A]. Actually, song [A] seems to be one of the more slower songs in the corpus, yet spotify has decided it has a high BPM. The chordograms also do not show any irregularities.

Novelty differences between Jazz Songs

On the top you can see self similarity matrices of a song from the elevator jazz list and a song from the general jazz list, on the bottom you can see the related novelty functions. I believe these two songs are good indicators of their respective playlists.

What is notable is that some repetition (dioganal lines) can be spotted in the beginning of “Night and Day”, after listening to the rest of the playlist occasional repetition can indeed be heard.

Au contraire, the general jazz song shows a lot of novelty (crosses). This song in particular keeps changing tempo, instruments, chords etc. This novelty is also illustrated by the high and frequent peaks of the novelty function.

Thus these self similarity matrices are quite characteristic of the playlist they represent. It can be concluded that the general jazz uses more novelty.

Conclusions: Why my jazz listening habits are the way they are.

This research led to some very interesting insights into my spotify playlist and my music listening habits.

The elevator jazz list has some very distinct characteristics, it has a extremely high acousticness and instrumentalness and a low energy. In opposition, the general jazz list has a very wide range of features. The biggest difference between the elevator jazz and its originals are a lack of vocals.

I believe it is easy to explain why I use the elevator jazz list mostly as background music, or when I am studying. A combination of a extremely high instrumentalness (no vocals), a low energy level and minimal novelty make this music enjoyable but not distracting. When I want to listen to something more “complex”, I have a wide range of other jazz to choose from.